Performance analysis of support vector machine based classifiers

Classification is a challenging problem in the various fields of knowledge i.e., Pattern Recognition, Data Mining, Knowledge Discovery from Database etc. There is various classification methods are proposed in the contemporary literature. The choice of an appropriate classifier to achieve the optimal performance on a specific problem needs more empirical studies. There are various algorithmic paradigms like, Associative Classification; Decision Trees based classification, Statistical Classification and Support Vector Machines etc. which are exploited for the classification purposes. This paper investigates the performance of Support Vector Machine (SVM) based classifiers namely SMO-C, C-SVM-C, and NU-SVM-C. SVM is a very successful classification approach for the binary classification as well as non-binary classification problems. This study, performance comparative analysis of SVM based classification approach on public data sets; exploit the implementation of the corresponding classifiers in the KEEL. The SVM-C approach wins one time, draw 5 times and lost 6 times with respective other approaches. The NU_SVM-C win one time, draw 4 times and lost 7 times while SMO-C wins 5 times, draw 3 times and loss 4 times. It is shown that the performance of SMO-C is promising with respect to other SVM based classifiers. (c) 2018 The Authors. Published by IASE.